Hearth lets you run AI agents on your home computer and manage them from your iPhone. The agents do the work; you stay in control of what they're allowed to do. This guide walks you through setup from scratch.
Open a terminal and run:
curl -sSL https://hearthcmd.dev/install.sh | bash
This downloads and installs the hearth command to /usr/local/bin/hearth. Confirm it's installed correctly:
which hearth
hearth -v
hearth login you@example.com
Hearth sends a 6-digit code to your email. Paste it at the prompt:
A 6-digit code has been sent to you@example.com.
Enter the code: 123456
On first login you'll be asked for your name and a name for your household (your "organization" in Hearth). Defaults are fine — just press Enter. Once verified, this computer is enrolled and ready.
The daemon is a small background process that keeps your computer connected and manages your agents:
hearth host start
Check that it's running:
hearth host status
To stop it: hearth host stop. Note that it does not restart automatically after a reboot — you'll need to run hearth host start again, or set it up as a startup item.
Open the Hearth app, enter the same email address, and verify with the code that arrives. The app will find the computer you just enrolled and show it connected.
Agents are the AI assistants that do the actual work. You can start one from the app or from your terminal.
Navigate to the folder you want the agent to work in, then:
hearth hh agent create --temp
This spins up a temp agent in the current directory and opens a live view of what it's doing.
When an agent wants to do something that needs your sign-off — read a file outside its working folder, access the web, run a system command — you'll get a notification. Open it to see exactly what the agent is asking to do, then tap:
The first few sessions will have more requests as the system learns your preferences. After you've tapped "Always Allow" a few times on the things you trust, most actions will happen automatically.
Every "Always Allow" creates a rule. Hearth's rules are specific: approving one web request doesn't open the floodgates to all web requests. Rules are scoped by action type, and for sensitive operations like deleting files, each exact command gets its own rule.
You can review, edit, and delete your rules any time in the Grants tab for any agent.
See the support page if something isn't working.
Email support@vergelabs.org — we're happy to help.